You're viewing a single thread.
Yes Nvidia is still a shitty company like every other. But their open source drivers run pretty well by now.
Assuming you're talking about Nouveau, it's pretty hit or miss depending on what card you have. My previous laptop had an MX330 and it couldn't do hardware acceleration stuff and 120Hz via HDMI, not to mention screen sharing on Wayland was wonky.
Oh, and it's worth to mention that "their" open source driver had nothing to do with Nvidia themselves; they absolute do not care, as opposed to AMD.
I think he means nvk. It's a whole new world. I thought I just heard it's ready. So worth checking out I guess.
Judging by the docs at mesa3d, I don't even have a card that supports it
NVK currently supports Turing (RTX 20XX and GTX 16XX) and later GPUs. Eventually, we plan to support as far back as Kepler (GeForce 600 and 700 series) GPUs but anything pre-Turing is currently disabled by default.
I'm not talking about Nouveau. I'm talking about the Open Source drivers from Nvidia. They released them a while ago and they've gotten pretty good lately.
I heard back than that the open source driver is pretty much just a layer of abstraction and that most actual code runs in the firmware which is proprietary
The driver you're likely referring to is NVK, which is also not developed by Nvidia, check out the annoucement post by Collabora, it says:
As said above, NVK is a new open-source Vulkan driver for NVIDIA hardware in Mesa. It's been written almost entirely from scratch using the new official headers from NVIDIA. We occasionally reference the existing nouveau OpenGL driver [...]
And also
a few months ago, NVIDIA released an open-source version of their kernel driver. While this isn't quite documentation, it does give us something to reference to see how NVIDIA drives their hardware. The code drop from NVIDIA isn't a good fit for upstream Linux, but it does give us the opportunity to rework the upstream driver situation and do it right.
So they're developing a driver based off headers made available by Nvidia and some of the reverse engineered code from regular Nouveau. In fact, it seems to be a branch of Nouveau as it stands:
Trying out NVK is no different than any other Mesa driver. Just pull the branch nvk/main branch from the nouveau/mesa project, build it, and give it a try
So the "OSS drivers from Nvidia" aren't what makes it work, it's the whole community effort to build NVK from scratch.
Regardless, it only supports the most recent cards using Turing architecture. From mesa3d docs:
NVK currently supports Turing (RTX 20XX and GTX 16XX) and later GPUs. Eventually, we plan to support as far back as Kepler (GeForce 600 and 700 series) GPUs but anything pre-Turing is currently disabled by default.
And still I'm not taking about nvk nor am I taking about nouveau. I'm taking about the open source drivers from nvidia. https://github.com/NVIDIA/open-gpu-kernel-modules https://developer.nvidia.com/blog/nvidia-releases-open-source-gpu-kernel-modules/
The kernel driver isn't all you need, you still have to pair it to their closed-source drivers or Nouveau/NVK, which are the device drivers
They released them a while ago and they've gotten pretty good lately.
Based on what?
Based on what I heard from people who use it. I don't have a nvidia card so can't tell.
The narrative around Nvidia seems to have done a complete 180 in the last year or two. I'm skeptical that it's as good as many are stating.
Are these OSS drivers maintained by Nvidia? Or is it Nouveau?
I know practically nothing about Nvidia's drivers, but I can see them doing a 180° flip, because they want to grab a chunk of that AI market.
Even before LLMs, they were investing there, and now it's just completely settled that tons of AI-related hardware needs to be either a beefy a server machine or a low-profile edge PC. For both, Linux is very much preferrable.Nvidia pretty much always dominated the AI GPU market with their closed source driver and Cuda. Nothing has changed about that except for more competition in AI specific hardware which you can buy from several vendors now. But no one has ever used AMD cards with OpenCL for AI or ML. If you were serious about it, you always used Nvidia with Cuda or nowadays some dedicated AI accelerator card (DPUs).
Yes, and with the addition of NVK it's gotten a lot better. There is still some issues on Wayland and specific problem cards, but it's not nearly as bad as it was even 2y ago.